Image source: ftp://public.dhe.ibm.com/software/analytics/spss/documentation/modeler/18.0/en/ModelerCRISPDM.pdf

image-2.png

Data Source¶

https://www.kaggle.com/datasets/airbnb/seattle

https://www.kaggle.com/datasets/airbnb/boston

The following Airbnb activity is included in this dataset:

  1. Listings, including full descriptions and average review score
  2. Reviews, including unique id for each reviewer and detailed comments
  3. Calendar, including listing id and the price and availability for that day

Import Library¶

In [4]:
import pandas as pd
import polars as pl
import numpy as np
import matplotlib.pyplot as plt
import sweetviz as sv
from io import StringIO
import plotly.express as px

import warnings
warnings.filterwarnings('ignore')

Gather Dataset¶

In [ ]:
!kaggle datasets download -d airbnb/seattle
!kaggle datasets download -d airbnb/boston

!unzip seattle.zip
!unzip boston.zip

Define utility function¶

Create helper function for data preprocessing for reusability.

In [5]:
def flag_city(df,city):
    "Add additional column 'city' with value as stated in parameter"
    df["city"]=city
    return df

def concat(df1,df2):
    "Union two dataframe into single dataframe using first dataframe columns as reference"
    return pd.concat([df1[df1.columns], df2[df1.columns]])

def to_date(df,columns):
    "Convert list of columns stated in parameter to date type"
    for col in columns:
        df[col] = pd.to_datetime(df[col]).dt.strftime('%Y-%m-%d')
    return df

def price_to_numeric(df,columns):
    """
    Cast price column in dataframe into numeric values.
    Remove $ sign & convert it to numeric
    """
    for col in columns:
        df[col] = pd.to_numeric(df[col].str.replace('$', '').str.replace(',', ''))
    return df

Read Data¶

Read each Boston & Seattle csv file into dataframe.
For each city there is 3 files, calendar, listing & reviews.

  1. Calendar

Contain property availability for specific period on daily level
2. Listings
Contain indepth atttribute of each listed property, such as price, location, facitilies, rating etc.
3. Review
Contain indepth review submitted on each listed property.

Explicitly read Id related column as string.

In [20]:
# Declare data type for each files
calendar_dtype = {'listing_id': str}
listing_dtype = {"id": str,"hos_id": str}
reviews_dtype = {"id": str, 'listing_id': str,'reviewer_id': str}

# Read csv in both City
seattle_calendar = flag_city(pd.read_csv("./seattle/calendar.csv",dtype= calendar_dtype),"Seattle")
seattle_listings = flag_city(pd.read_csv("./seattle/listings.csv",dtype= listing_dtype),"Seattle")
seattle_reviews = flag_city(pd.read_csv("./seattle/reviews.csv",dtype= reviews_dtype),"Seattle")

boston_calendar = flag_city(pd.read_csv("./boston/calendar.csv",dtype= calendar_dtype),"Boston")
boston_listings = flag_city(pd.read_csv("./boston/listings.csv",dtype= listing_dtype),"Boston")
boston_reviews = flag_city(pd.read_csv("./boston/reviews.csv",dtype= reviews_dtype),"Boston")

In this cell we do following data preprocessing:

  1. Union Seattle & Bostong into single dataframe
  2. Cast date related column to date data type
  3. Parse price into numeric data type
  4. Update superhost column values from 't' or 'f' into 'Superhost' or 'Non-Superhost'.

Data Cleansing & Preprocessing¶

In [27]:
# Union Calendar
calendar = pd.concat([seattle_calendar, boston_calendar])
calendar = to_date(calendar,["date"])
calendar = price_to_numeric(calendar,["price"])

# Neighbour group column in Boston are empty, we will replace it with cleansed neighbour column instead 
boston_listings["neighbourhood_group_cleansed"] = boston_listings["neighbourhood_cleansed"]
# Union Listing
listings = pd.concat([seattle_listings, boston_listings])
listings = to_date(listings,["host_since","first_review","last_review"])
listings = price_to_numeric(listings,["price","weekly_price","monthly_price","cleaning_fee"])

# Update superhost column value
superhost_cond = (
    listings["host_is_superhost"] == 't',
    listings["host_is_superhost"] == 'f',
)
listings["host_is_superhost"] = np.select(superhost_cond, ['Superhost', 'Non-Superhost'], default=None)

listings[listings["host_is_superhost"]=='t']["host_is_superhost"] = 'Superhost'
listings[listings["host_is_superhost"]=='f']["host_is_superhost"] = 'Non-Superhost'

# Union reviews columns
reviews = pd.concat([seattle_reviews, boston_reviews])
reviews = to_date(reviews,["date"])

Sanity Check¶

Data sanity check to see:

  1. Number of record in calendar data for each city
  2. When is first & last observation in calendar
  3. Number of hoster & listing
In [45]:
print(f"""
    Seattle Data
    Number of observation: {np.shape(calendar[calendar["city"]=="Seattle"])[0]} records
    Fist observation: {str(calendar[calendar["city"]=="Seattle"]["date"].min())}
    Last observation: {str(calendar[calendar["city"]=="Seattle"]["date"].max())}
    
    Total Host: {np.shape(listings[listings["city"]=="Seattle"]["host_id"].unique())[0]}
    Total Listing: {np.shape(listings[listings["city"]=="Seattle"]["id"].unique())[0]}

""")


print(f"""
    Boston Data
    Number of observation: {np.shape(calendar[calendar["city"]=="Boston"])[0]} records
    Fist observation: {str(calendar[calendar["city"]=="Boston"]["date"].min())}
    Last observation: {str(calendar[calendar["city"]=="Boston"]["date"].max())}
    
    Total Host: {np.shape(listings[listings["city"]=="Boston"]["host_id"].unique())[0]}
    Total Listing: {np.shape(listings[listings["city"]=="Boston"]["id"].unique())[0]}
""")

    Seattle Data
    Number of observation: 1393570 records
    Fist observation: 2016-01-04
    Last observation: 2017-01-02
    
    Total Host: 2751
    Total Listing: 3818



    Boston Data
    Number of observation: 1308890 records
    Fist observation: 2016-09-06
    Last observation: 2017-09-05
    
    Total Host: 2181
    Total Listing: 3585

EDA with Sweetviz Library¶

To speed up process we use python library Sweetviz to analyze 3 data (calendar, listings & reviews).
The result should are following:

  1. calendar.html
  2. listings.html
  3. reviews.html
In [12]:
analyze_report = sv.analyze(calendar)
analyze_report.show_html("calendar.html",open_browser=True)
In [13]:
analyze_report = sv.analyze(listings)
analyze_report.show_html("listings.html",open_browser=True)
In [14]:
analyze_report = sv.analyze(reviews)
analyze_report.show_html("reviews.html",open_browser=True)

Q1. How is property distribution listed in Seattle compared to Boston?¶

We answer this question by analyzing listing dataframe particularly price column.
If look at price statistic & visualizing price distribution in both in both city, we can see that in general Boston are more expensive than seattle as shown in statistics.As for price distribution both Seattle & Boston distribution are centered around 85-90 USD. The difference is Boston property have fatter right tail which show more option for high price property.

In [167]:
df[df["city"]=="Boston"]["price"].describe()
Out[167]:
count    2748.000000
mean      167.188137
std       113.097098
min        11.000000
25%        85.000000
50%       146.000000
75%       215.000000
max      1300.000000
Name: price, dtype: float64
In [168]:
df[df["city"]=="Seattle"]["price"].describe()
Out[168]:
count    3166.000000
mean      126.844599
std        90.068497
min        22.000000
25%        75.000000
50%       100.000000
75%       150.000000
max      1000.000000
Name: price, dtype: float64
In [8]:
import plotly.figure_factory as ff

# For visualization clarity, we limit our distribution analysis for property price below USD 1500
df = listings[listings["price"]<1500]

hist_data = [df[df["city"]=="Boston"]["price"], df[df["city"]=="Seattle"]["price"]]

group_labels = ['Boston', 'Seattle']

# Create distplot
fig = ff.create_distplot(
    hist_data, group_labels,
    show_hist=False,
)

fig.update_layout(title_text='Listing Price Distribution',template='simple_white',xaxis_title="Price",)
fig.show()

Q2. Is there any pricing difference for particular property type between those cities?¶

Next we are checking property type structure in both city and then compare the pricing for the same property type.

We can see that Seattle listing are predominantly consist of house and apartment with each more than 1.7K listings, while in Boston majority apartment are the majority listing, ~2.6K and house significantly smaller ~0.5K. Boston also have more luxurious property such as Townhouse & Condominium compared to Seattle.

To compare price we use boxplot to analyze dominant property type (House,Apartment,Condominium,Townhouse, Bed & Breakfast,Loft). Aside from house, Boston have higher price in all dominant property mentioned above.

In [10]:
import plotly.express as px
from plotly.subplots import make_subplots

# Filter City
df_seattle = listings[listings["city"]=="Seattle"]
# Aggregate=ing dataframe
df_seattle = (
    df_seattle[["city","property_type","id","price"]]
    .groupby(['city','property_type'])
    .agg({'id':'size', 'price':'mean'}) 
    .reset_index()
).sort_values(by='id', ascending=True)

# Create Barplot
fig_seattle = px.bar(df_seattle,
             y="property_type", x="id",color="price",text="price",
             orientation='h'
).update_layout(
    title_text='Seattle Property Type',template='simple_white',
    yaxis={'categoryorder':'total ascending'},
    yaxis_title="Property Type",
    xaxis_title="Number of Listing",
).update_traces(texttemplate='%{text:.2s}')

# Filter City
df_boston = listings[listings["city"]=="Boston"]
# Aggregate=ing dataframe
df_boston = (
    df_boston[["city","property_type","id","price"]]
    .groupby(['city','property_type'])
    .agg({'id':'size', 'price':'mean'}) 
    .reset_index()
).sort_values(by='id', ascending=True)

# Create Barplot
fig_boston = px.bar(df_boston,
             y="property_type", x="id",color="price",text="price",
             orientation='h'
).update_layout(
    title_text='Boston Property Type',template='simple_white',
    yaxis={'categoryorder':'total ascending'},
    yaxis_title="Property Type",
    xaxis_title="Number of Listing",
).update_traces(texttemplate='%{text:.2s}')

# Prepare tracer for subplot
figure1_traces = []
figure2_traces = []

for trace in range(len(fig_seattle["data"])):
    figure1_traces.append(fig_seattle["data"][trace])
for trace in range(len(fig_boston["data"])):
    figure2_traces.append(fig_boston["data"][trace])

fig_subplot = make_subplots(rows=1, cols=2, subplot_titles=('Seattle',  'Boston'))
for traces in figure1_traces:
    fig_subplot.append_trace(traces, row=1, col=1)
for traces in figure2_traces:
    fig_subplot.append_trace(traces, row=1, col=2)

# Update subplot layout config
fig_subplot.update_layout(
    title_text='Property Type (color indicate price)',template='simple_white',
    yaxis_title="Property Type",
    xaxis_title="Number of Listing",
    showlegend=True,
).update_traces(texttemplate='%{text:.2s}')

fig_subplot['layout']['xaxis']['title']='Number of Listing'
fig_subplot['layout']['xaxis2']['title']='Number of Listing'

fig_subplot.show()
In [9]:
majority_type = ["House","Apartment","Condominium","Townhouse", "Bed & Breakfast","Loft"]
fig = px.box(listings[listings["property_type"].isin(majority_type)],
             y="property_type", x="price",color="city",
             orientation='h'
)
fig.update_layout(
    title_text='Price by Type',template='simple_white',
    yaxis_title="Type",
    xaxis_title="Price",
    showlegend=True,
)
fig.update_layout(xaxis_range=[0,800])
fig.show()

Q3. How is the average rating differ between those cities, by property type?¶

Given it higher pricing, does Boston customer have higher customer satisfaction compared to Seattle, or is it the opposite?

It turns out, Boston have lower satisfaction in general compared to Seattle, this can be seen from distplot showing Boston have fatter left tail on customer review rating. Moreover if we look on property type level, we can see that Seattle consistenly have higher satisfaction across all property type.

In [11]:
import plotly.figure_factory as ff

df = listings[(listings["review_scores_rating"].notnull()) & (listings["review_scores_rating"]> 50)]

hist_data = [df[df["city"]=="Boston"]["review_scores_rating"], df[df["city"]=="Seattle"]["review_scores_rating"]]

group_labels = ['Boston', 'Seattle']

# Create distplot with curve_type set to 'normal'
fig = ff.create_distplot(
    hist_data, group_labels,
    show_hist=False,
)

fig.update_layout(title_text='Overall Rating Distribution',template='simple_white',xaxis_title="Review Score",)
fig.show()
In [12]:
import plotly.express as px
from plotly.subplots import make_subplots

# Filter & Aggregate dataframe
df_seattle = listings[listings["city"]=="Seattle"]
df_seattle = (
    df_seattle[["city","property_type","id","review_scores_rating","number_of_reviews"]]
    .groupby(['city','property_type'])
    .agg({'id':'size', 'review_scores_rating':'mean',"number_of_reviews": "sum"}) 
    .reset_index()
).sort_values(by='review_scores_rating', ascending=True)

# Create Barplot
fig_seattle = px.bar(df_seattle,
             y="property_type", x="review_scores_rating",color="number_of_reviews",text="review_scores_rating",
             orientation='h'
).update_layout(
    title_text='Seattle Property Type',template='simple_white',
    yaxis={'categoryorder':'total ascending'},
    yaxis_title="Property Type",
    xaxis_title="Number of Listing",
).update_traces(texttemplate='%{text:.2s}')

# Filter & Aggregate dataframe
df_boston = listings[listings["city"]=="Boston"]
df_boston = (
    df_boston[["city","property_type","id","review_scores_rating","number_of_reviews"]]
    .groupby(['city','property_type'])
    .agg({'id':'size', 'review_scores_rating':'mean',"number_of_reviews": "sum"}) 
    .reset_index()
).sort_values(by='review_scores_rating', ascending=True)
df_boston = df_boston[df_boston["review_scores_rating"].notnull()]

# Create Barplot
fig_boston = px.bar(df_boston,
             y="property_type", x="review_scores_rating",color="number_of_reviews",text="review_scores_rating",
             orientation='h'
).update_layout(
    title_text='Boston Property Type',template='simple_white',
    yaxis={'categoryorder':'total ascending'},
    yaxis_title="Property Type",
    xaxis_title="Number of Listing",
).update_traces(texttemplate='%{text:.2s}')

# Prepare subplot
figure1_traces = []
figure2_traces = []

for trace in range(len(fig_seattle["data"])):
    figure1_traces.append(fig_seattle["data"][trace])
for trace in range(len(fig_boston["data"])):
    figure2_traces.append(fig_boston["data"][trace])

fig_subplot = make_subplots(rows=1, cols=2, subplot_titles=('Seattle',  'Boston'),horizontal_spacing = 0.15)
for traces in figure1_traces:
    fig_subplot.append_trace(traces, row=1, col=1)
for traces in figure2_traces:
    fig_subplot.append_trace(traces, row=1, col=2)

fig_subplot.update_layout(
    title_text='Review by Type (color indicate number of review)',template='simple_white',
    yaxis_title="Property Type",
    xaxis_title="Number of Listing",
    showlegend=True,
)

fig_subplot['layout']['xaxis']['title']='Average Rating'
fig_subplot['layout']['xaxis2']['title']='Average Rating'

fig_subplot.show()

Q4. Is higher score due to lower price?¶

So, maybe higher satisfaction level in Seattle is due to its lower price?

Based on scatterplot visualiztion, there is no indication of strong non-linearity between price & rating, so we will use pearson standard correlation for simplicity.
If we look at correlation result, we can see both in Boston & Seattle price doesnt really correlate with price (11th in Seattle & 14th in Boston). Surpisingly top 5 factor that correlate to satisfaction in both city are in hospitaly & hoster transaprency qualities, such as cleanliness, trustworthyness (accuracy), checkin experience & communication. It turns out high review rating are mainly about fulfilling customer expectation.

Seattle Correlation Result¶

In [261]:
correlation = listings[listings["city"]=="Seattle"].corr().reset_index()
correlation["absolute_corr"] = np.abs(correlation["review_scores_rating"]).fillna(0)
correlation = correlation.sort_values(by=['absolute_corr'], ascending=False)[["index","review_scores_rating"]].reset_index(drop=True)
correlation = correlation.rename(columns={"index": "factor"})
correlation = correlation.loc[2:15,:].reset_index(drop=True)
correlation.index += 1
correlation
Out[261]:
factor review_scores_rating
1 review_scores_cleanliness 0.642882
2 review_scores_accuracy 0.621257
3 review_scores_communication 0.540620
4 review_scores_checkin 0.521813
5 review_scores_location 0.368423
6 calculated_host_listings_count -0.219280
7 square_feet 0.143793
8 host_listings_count -0.109357
9 host_total_listings_count -0.109357
10 reviews_per_month 0.087313
11 price 0.055551
12 availability_30 -0.051439
13 monthly_price 0.048595
14 availability_60 -0.048460

Scatterplot visualization to check if there is strong non-linearity betwen price & rating.¶

There is no indication of strong non-linearity between price & rating.

In [13]:
fig = px.scatter(listings, x="review_scores_rating", y="price", color="city")
fig.update_layout(yaxis_range=[0,1000])
fig.show()

Boston Correlation Result¶

In [260]:
correlation = listings[listings["city"]=="Boston"].corr().reset_index()
correlation["absolute_corr"] = np.abs(correlation["review_scores_rating"]).fillna(0)
correlation = correlation.sort_values(by=['absolute_corr'], ascending=False)[["index","review_scores_rating"]].reset_index(drop=True)
correlation = correlation.rename(columns={"index": "factor"})
correlation = correlation.loc[2:15,:].reset_index(drop=True)
correlation.index += 1
correlation
Out[260]:
factor review_scores_rating
1 review_scores_cleanliness 0.754327
2 review_scores_accuracy 0.705104
3 review_scores_communication 0.600599
4 review_scores_checkin 0.584034
5 review_scores_location 0.458901
6 square_feet -0.175922
7 availability_30 -0.147676
8 calculated_host_listings_count -0.142220
9 availability_60 -0.138288
10 host_listings_count -0.125055
11 host_total_listings_count -0.125055
12 availability_90 -0.123666
13 weekly_price 0.109238
14 price 0.105651

Q5 Is some there particular neighbour with high rating?¶

In both Seattle & Boston customer rating are relatively similar across the neighbour. With only noticable lower rating (score less than 90) in Seattle's University district & Boston's Mission Hill.

In [18]:
import plotly.express as px
from plotly.subplots import make_subplots

# Filter & group dataframe
df_seattle = listings[listings["city"]=="Seattle"]
df_seattle = (
    df_seattle[["city","neighbourhood_group_cleansed","review_scores_rating","number_of_reviews"]]
    .groupby(['city','neighbourhood_group_cleansed'])
    .agg({'review_scores_rating':'mean',"number_of_reviews": "sum"}) 
    .reset_index()
).sort_values(by='review_scores_rating', ascending=True)
df_seattle = (
    df_seattle[df_seattle["review_scores_rating"].notnull()]
    .sort_values(by='review_scores_rating', ascending=True)
)

# Create Barplot
fig_seattle = px.bar(df_seattle,
             y="neighbourhood_group_cleansed", x="review_scores_rating",color="number_of_reviews",text="review_scores_rating",
             orientation='h'
).update_layout(
    title_text='Seattle Property Type',template='simple_white',
    yaxis={'categoryorder':'total ascending'},
    yaxis_title="Property Type",
    xaxis_title="Number of Listing",
).update_traces(texttemplate='%{text:.2s}')

# Filter & group dataframe
df_boston = listings[listings["city"]=="Boston"]
df_boston = (
    df_boston[["city","neighbourhood_group_cleansed","review_scores_rating","number_of_reviews"]]
    .groupby(['city','neighbourhood_group_cleansed'])
    .agg({'review_scores_rating':'mean',"number_of_reviews": "sum"}) 
    .reset_index()
).sort_values(by='review_scores_rating', ascending=True)
df_boston = (
    df_boston[df_boston["review_scores_rating"].notnull()]
    .sort_values(by='review_scores_rating', ascending=True)
)

# Create Barplot
fig_boston = px.bar(df_boston,
             y="neighbourhood_group_cleansed", x="review_scores_rating",color="number_of_reviews",text="review_scores_rating",
             orientation='h'
).update_layout(
    title_text='Boston Property Type',template='simple_white',
    yaxis={'categoryorder':'total ascending'},
    yaxis_title="Property Type",
    xaxis_title="Number of Listing",
).update_traces(texttemplate='%{text:.2s}')

# Subplotting
figure1_traces = []
figure2_traces = []

for trace in range(len(fig_seattle["data"])):
    figure1_traces.append(fig_seattle["data"][trace])
for trace in range(len(fig_boston["data"])):
    figure2_traces.append(fig_boston["data"][trace])

fig_subplot = make_subplots(rows=1, cols=2, subplot_titles=('Seattle',  'Boston'),horizontal_spacing = 0.22)
for traces in figure1_traces:
    fig_subplot.append_trace(traces, row=1, col=1)
for traces in figure2_traces:
    fig_subplot.append_trace(traces, row=1, col=2)

fig_subplot.update_layout(
    title_text='Review by Type (color indicate number of review)',template='simple_white',
    yaxis_title="Property Type",
    xaxis_title="Number of Listing",
    showlegend=True,
)

fig_subplot['layout']['xaxis']['title']='Average Rating'
fig_subplot['layout']['xaxis2']['title']='Average Rating'

fig_subplot.show()

Superhost Eligibility Criteria

image.png

Q6. Is there any difference in term of price betwen Superhost vs Non-Superhost?¶

  1. Price

For majority of segment such as apartment & house, pricing are relatively similar. Although there is some exception such as Seattle’s townhouse & condominium where price spread are more wide & price median is higher on Superhost.

2. Rating
As expected Superhost have more consistent rating, which can be seen seen in smaller standard deviation & higher minimum rating column. This is expected since they have obligation to maintain overall rating and keep cancellation rate below 1%.

In [19]:
# For visual clarity in we will exclude property price above 800
# Only analyze listing with superhost column not null
df_seattle = listings[
    (listings["host_is_superhost"].notnull()) &
    (listings["city"] == "Seattle") &
    (listings["price"] < 800)
]

majority_type = ["House","Apartment","Condominium","Townhouse", "Bed & Breakfast","Loft"]
fig_seattle = px.box(df_seattle[df_seattle["property_type"].isin(majority_type)],
             y="property_type", x="price",color="host_is_superhost",
             orientation='h',
)
fig_seattle.update_layout(
    title_text='Price by Type',template='simple_white',
    yaxis_title="Type",
    xaxis_title="Price",
)

# For visual clarity in we will exclude property price above 800
# Only analyze listing with superhost column not null
df_boston = listings[
    (listings["host_is_superhost"].notnull()) &
    (listings["city"] == "Boston") &
    (listings["price"] < 800)
]

majority_type = ["House","Apartment","Condominium","Townhouse", "Bed & Breakfast","Loft"]
fig_boston = px.box(df_boston[df_boston["property_type"].isin(majority_type)],
             y="property_type", x="price",color="host_is_superhost",
             orientation='h'
)
fig_boston.update_layout(
    title_text='Price by Type',template='simple_white',
    yaxis_title="Type",
    xaxis_title="Price",
    showlegend=False,
)

# Subplotting
figure1_traces = []
figure2_traces = []

for trace in range(len(fig_seattle["data"])):
    figure1_traces.append(fig_seattle["data"][trace])
for trace in range(len(fig_boston["data"])):
    figure2_traces.append(fig_boston["data"][trace])

fig_subplot = make_subplots(rows=1, cols=2, subplot_titles=('Seattle',  'Boston'),horizontal_spacing = 0.15,vertical_spacing = 1)
for traces in figure1_traces:
    fig_subplot.append_trace(traces, row=1, col=1)
for traces in figure2_traces:
    fig_subplot.append_trace(traces, row=1, col=2)

fig_subplot.update_layout(
    title_text='Price Superhost vs Non-Superhost',template='simple_white',
    yaxis_title="Type",
    xaxis_title="Number of Listing",
    boxmode='group',
)

fig_subplot['layout']['xaxis']['title']='Price'
fig_subplot['layout']['xaxis2']['title']='Price'
fig_subplot.update_layout(legend=dict(
    orientation="h",
))

fig.update_layout(
   title="My plot",
   xaxis_title="id",
   yaxis_title="salary",
   legend_title="legend",
   font=dict(family="Arial", size=20, color="green")
)
fig_subplot.show()

To compare multiple rating, we simply use pandas describe function

In [345]:
df = listings[listings["city"]=="Seattle"]
df[['host_is_superhost',
         'review_scores_rating','review_scores_accuracy', 'review_scores_cleanliness',
         'review_scores_checkin', 'review_scores_communication',
         'review_scores_location', 'review_scores_value'
]].groupby(['host_is_superhost']).describe().T.unstack(1)
Out[345]:
host_is_superhost Non-Superhost Superhost
count mean std min 25% 50% 75% max count mean std min 25% 50% 75% max
review_scores_rating 2420.0 93.652893 7.192689 20.0 91.0 95.0 100.0 100.0 751.0 97.395473 2.624384 60.0 96.0 98.0 99.0 100.0
review_scores_accuracy 2409.0 9.552511 0.760750 2.0 9.0 10.0 10.0 10.0 751.0 9.905459 0.314722 8.0 10.0 10.0 10.0 10.0
review_scores_cleanliness 2414.0 9.454018 0.864743 3.0 9.0 10.0 10.0 10.0 751.0 9.885486 0.365425 6.0 10.0 10.0 10.0 10.0
review_scores_checkin 2409.0 9.731424 0.663235 2.0 10.0 10.0 10.0 10.0 751.0 9.964048 0.200098 8.0 10.0 10.0 10.0 10.0
review_scores_communication 2416.0 9.758278 0.635341 2.0 10.0 10.0 10.0 10.0 751.0 9.974700 0.165406 8.0 10.0 10.0 10.0 10.0
review_scores_location 2412.0 9.564677 0.661469 4.0 9.0 10.0 10.0 10.0 751.0 9.750999 0.485025 6.0 10.0 10.0 10.0 10.0
review_scores_value 2411.0 9.359602 0.795533 2.0 9.0 9.0 10.0 10.0 751.0 9.749667 0.471787 6.0 10.0 10.0 10.0 10.0

Enable plotly offline mode before saving to html

In [28]:
import plotly
plotly.offline.init_notebook_mode()
In [ ]:
In [ ]: